Many years ago I developed a similar system to that developed
by Steve Skiena
in his book "Calculated Bets".
His book is the definitive book on how to beat jai-alai.
I differ with Skiena on some minor points and, maybe, one major point. The major point is that he appears to use a longer timeline in computing his ratings. I prefer to use a short timeline with additional weighting the more recent the performance.
I have "manually" done my ratings over the years. I have not done them in recent time because of time restrictions.
My process is as follows for singles games:
(1) I print the entries
(2) I print the results
(3) I input into a spread sheet the player pf/pa rating for that game.
For example, if a player wins the game from post 8, I will input into
the spreadsheet the player rating of 400 and 55 into separate cells.
The 400 stands for 4.00 and the 55 stands for 0.55. This indicates that
this player and post, on average, will score 4.0 times and lose 0.55
times when winning from post 8. I do this for all posts.
(4) I then total up all the "points for" and "points against" for that
player in recent performances.
(5) From this I compute a player rating with 500 indicating an average rating.
(6) I then input this rating into a Qbasic program and determine the
various performance probabilites of the players depending upon which
post they are playing from.
(7) I print out these performan ce probabilities in the form of an
odds line for Win and Exactas and quinellas
This is a long tedious process and can be better automated like Steve
Skiena does in his book even to the point where his wagers are placed automatically.
I am sure that Tiger and others are using a similar system whose ratings
vary from mine to the degree that they are using a different timeline
and different weightings.
Does anyone know how my system can be better automated ? Particularly,
the input factor for each player that requires a 2-fold Points For
and Points Against input that may not be amenable to a direct automated DB entry